You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Copy file name to clipboardExpand all lines: _news/consultancy.md
+8-9Lines changed: 8 additions & 9 deletions
Display the source diff
Display the rich diff
Original file line number
Diff line number
Diff line change
@@ -6,7 +6,6 @@ related_posts: false
6
6
title: New Launch - Robotics & Automation Consultancy
7
7
---
8
8
9
-
10
9
**🚀 New Launch: Robotics & Automation Consultancy**
11
10
12
11
I’m excited to announce the launch of my **Consultancy Service** for individuals, startups, and organizations working on **robotics and automation** projects.
@@ -15,14 +14,14 @@ Whether you're stuck on embedded systems, need help debugging, or want guidance
Copy file name to clipboardExpand all lines: _news/fluxpointv2.md
+49-48Lines changed: 49 additions & 48 deletions
Display the source diff
Display the rich diff
Original file line number
Diff line number
Diff line change
@@ -16,86 +16,88 @@ This version marks a major overhaul of the `fluxpoint.py` library, aligning it c
16
16
17
17
### ✨ Highlights
18
18
19
-
* ✅ **Modular Restructuring**: The entire library has been split into clearly defined modules based on endpoint categories such as `Convert`, `Color`, `Minecraft`, `Utility`, `ImageGen`, and more.
20
-
* ⚙️ **Complete API Coverage**: The wrapper now supports nearly all documented endpoints including:
21
-
22
-
*`/convert`
23
-
*`/color`
24
-
*`/mc`
25
-
*`/utility`
26
-
*`/image-gen/custom-image`
27
-
*`/image-gen/templates`
28
-
*`/sfw` and `/nsfw` images & gifs
29
-
* 📦 **Dynamic Versioning**: The version is now updated internally and reflected dynamically in requests.
30
-
* 📚 **Rebranded Documentation**:
31
-
32
-
* Moved under the new namespace: `Creatrix-Net`
33
-
* ReadTheDocs link updated: [[fluxpointpy.dhruvashaw.in](https://fluxpointpy.dhruvashaw.in/en/latest/)](https://fluxpointpy.dhruvashaw.in/en/latest/)
19
+
- ✅ **Modular Restructuring**: The entire library has been split into clearly defined modules based on endpoint categories such as `Convert`, `Color`, `Minecraft`, `Utility`, `ImageGen`, and more.
20
+
- ⚙️ **Complete API Coverage**: The wrapper now supports nearly all documented endpoints including:
21
+
22
+
-`/convert`
23
+
-`/color`
24
+
-`/mc`
25
+
-`/utility`
26
+
-`/image-gen/custom-image`
27
+
-`/image-gen/templates`
28
+
-`/sfw` and `/nsfw` images & gifs
29
+
30
+
- 📦 **Dynamic Versioning**: The version is now updated internally and reflected dynamically in requests.
31
+
- 📚 **Rebranded Documentation**:
32
+
33
+
- Moved under the new namespace: `Creatrix-Net`
34
+
- ReadTheDocs link updated: [[fluxpointpy.dhruvashaw.in](https://fluxpointpy.dhruvashaw.in/en/latest/)](https://fluxpointpy.dhruvashaw.in/en/latest/)
34
35
35
36
---
36
37
37
38
### 🔧 Changes & Improvements
38
39
39
40
#### 🧱 Structural
40
41
41
-
* Refactored project from monolithic classes to modular endpoint-driven structure.
42
-
* Introduced new directories: `paths/`, `vars.py`, etc.
43
-
* Deleted legacy files like `enums.py`, `images.py`, `gifs.py`, and `nsfw.py`.
42
+
- Refactored project from monolithic classes to modular endpoint-driven structure.
43
+
- Introduced new directories: `paths/`, `vars.py`, etc.
44
+
- Deleted legacy files like `enums.py`, `images.py`, `gifs.py`, and `nsfw.py`.
44
45
45
46
#### 🖼️ Image Generation
46
47
47
-
* Split `ImageGenerator` into:
48
+
- Split `ImageGenerator` into:
48
49
49
-
*`CustomImage`: For fully customizable graphics using `images`, `texts`, colors, dimensions.
50
-
*`Template`: For template-based welcome images.
50
+
-`CustomImage`: For fully customizable graphics using `images`, `texts`, colors, dimensions.
51
+
-`Template`: For template-based welcome images.
51
52
52
53
#### 🌈 New Features
53
54
54
-
***Color API**:
55
+
-**Color API**:
55
56
56
-
*`random()` – fetch random colors
57
-
*`info()` – fetch color info by name, hex, or RGB
57
+
-`random()` – fetch random colors
58
+
-`info()` – fetch color info by name, hex, or RGB
58
59
59
-
***Convert API**:
60
+
-**Convert API**:
60
61
61
-
* HTML ↔ Markdown
62
-
* Image format conversion (`png`, `webp`, `jpg`) with quality settings
62
+
- HTML ↔ Markdown
63
+
- Image format conversion (`png`, `webp`, `jpg`) with quality settings
63
64
64
-
***Minecraft API**:
65
+
-**Minecraft API**:
65
66
66
-
* Ping server, get player UUID, retrieve skins with `SkinType`
67
+
- Ping server, get player UUID, retrieve skins with `SkinType`
67
68
68
-
***Utility API**:
69
+
-**Utility API**:
69
70
70
-
* Convert Unix timestamp / Discord snowflake to human-readable formats
71
+
- Convert Unix timestamp / Discord snowflake to human-readable formats
71
72
72
-
***List API**:
73
+
-**List API**:
73
74
74
-
* Fetch lists of available banners, icons, fonts
75
+
- Fetch lists of available banners, icons, fonts
75
76
76
-
***Tests API**:
77
+
-**Tests API**:
77
78
78
-
* Provides endpoints to validate API, images, JSON, gallery, and error handling
79
+
- Provides endpoints to validate API, images, JSON, gallery, and error handling
79
80
80
81
---
81
82
82
83
### 🐛 Fixes & Minor Updates
83
84
84
-
* Fixed hardcoded API links in examples & README.
85
-
* Improved Windows compatibility with proper asyncio event loop policies.
86
-
* Updated installation instructions for Python ≥ 3.9.
87
-
* Cleaned up and replaced outdated or incorrect references and examples.
85
+
- Fixed hardcoded API links in examples & README.
86
+
- Improved Windows compatibility with proper asyncio event loop policies.
87
+
- Updated installation instructions for Python ≥ 3.9.
88
+
- Cleaned up and replaced outdated or incorrect references and examples.
88
89
89
90
---
90
91
91
92
### 🧪 Examples Refreshed
92
93
93
-
* Simplified and tested examples across:
94
+
- Simplified and tested examples across:
94
95
95
-
*`simple_request.py`
96
-
*`custom_generator_test.py`
97
-
*`welcome_image.py`
98
-
* Examples now match the refactored client interface.
96
+
-`simple_request.py`
97
+
-`custom_generator_test.py`
98
+
-`welcome_image.py`
99
+
100
+
- Examples now match the refactored client interface.
99
101
100
102
---
101
103
@@ -108,10 +110,9 @@ Updated to reflect all new classes, methods, and expected outputs. See the full
Copy file name to clipboardExpand all lines: _news/phishnet.md
+7-7Lines changed: 7 additions & 7 deletions
Display the source diff
Display the rich diff
Original file line number
Diff line number
Diff line change
@@ -13,18 +13,18 @@ PhishNet is designed to address the growing threat of sophisticated phishing att
13
13
14
14
Our solution is built on a robust, scalable architecture and a unique dual-model AI approach:
15
15
16
-
***Proactive Threat Identification**: We use a powerful fuzzy permutation method with the `dnstwist` package to generate and check for malicious domains in near real-time.
17
-
***Dual-Model AI Engine**: Our core is a pipeline of two **Random Forest Classifiers**. One model analyzes structural features of a URL, while the second, more accurate model, also incorporates the raw URL string itself to learn intricate patterns.
18
-
***Scalable Backend**: PhishNet is built using **Django** and **PostgreSQL** for a secure, scalable foundation. We handle all asynchronous tasks and user-submitted requests using **Celery** to ensure efficiency.
19
-
***Automated Reporting**: The system processes submitted URLs and automatically generates and emails detailed reports to the user.
16
+
-**Proactive Threat Identification**: We use a powerful fuzzy permutation method with the `dnstwist` package to generate and check for malicious domains in near real-time.
17
+
-**Dual-Model AI Engine**: Our core is a pipeline of two **Random Forest Classifiers**. One model analyzes structural features of a URL, while the second, more accurate model, also incorporates the raw URL string itself to learn intricate patterns.
18
+
-**Scalable Backend**: PhishNet is built using **Django** and **PostgreSQL** for a secure, scalable foundation. We handle all asynchronous tasks and user-submitted requests using **Celery** to ensure efficiency.
19
+
-**Automated Reporting**: The system processes submitted URLs and automatically generates and emails detailed reports to the user.
20
20
21
21
### Looking Ahead
22
22
23
23
This is just the beginning for PhishNet. We have a clear roadmap for future improvements, including:
24
24
25
-
* Integrating the `subfinder` module for advanced subdomain discovery.
26
-
* Performing historical data analysis using `cdx_toolkit`.
27
-
* Incorporating **LSTM networks** to further enhance our AI model's accuracy.
25
+
- Integrating the `subfinder` module for advanced subdomain discovery.
26
+
- Performing historical data analysis using `cdx_toolkit`.
27
+
- Incorporating **LSTM networks** to further enhance our AI model's accuracy.
28
28
29
29
We're excited to continue developing PhishNet and contribute to a safer digital landscape. For more information, please explore our project repository.
Copy file name to clipboardExpand all lines: _news/python-free-webinar.md
+4-3Lines changed: 4 additions & 3 deletions
Display the source diff
Display the rich diff
Original file line number
Diff line number
Diff line change
@@ -13,9 +13,10 @@ Whether you're a student, a professional looking to upskill, or simply an ambiti
13
13
This free session is a great way to experience the quality of our training and see how our comprehensive paid course can help you master Python and accelerate your career in automation, AI, and robotics.
14
14
15
15
**Webinar Details:**
16
-
***Date:** September 5th, 2025
17
-
***Time:** 20:00 - 22:00 IST
18
-
***Cost:** FREE
16
+
17
+
-**Date:** September 5th, 2025
18
+
-**Time:** 20:00 - 22:00 IST
19
+
-**Cost:** FREE
19
20
20
21
Secure your spot now and take the first step toward a future in tech!
Copy file name to clipboardExpand all lines: _pages/careers.md
+2Lines changed: 2 additions & 0 deletions
Display the source diff
Display the rich diff
Original file line number
Diff line number
Diff line change
@@ -7,6 +7,7 @@ nav: false
7
7
---
8
8
9
9
{% for job in site.data.careers %}
10
+
10
11
<divclass="job-listing">
11
12
<h3>{{ job.title }}</h3>
12
13
{% if job.department %}
@@ -19,6 +20,7 @@ nav: false
19
20
{% endfor %}
20
21
21
22
---
23
+
22
24
We are always looking for talented and passionate individuals to join our team. You can fill in the talent form below, and we will get back to you if there is a suitable opportunity.
Copy file name to clipboardExpand all lines: _projects/phishnet.md
+41-34Lines changed: 41 additions & 34 deletions
Display the source diff
Display the rich diff
Original file line number
Diff line number
Diff line change
@@ -31,23 +31,23 @@ Given this urgent and persistent threat, there is a critical need for a scalable
31
31
32
32
### Key Features
33
33
34
-
***Proactive Domain Identification**: Utilizes fuzzy permutation with the `dnstwist` library to actively find potential phishing domains before they are used in attacks.
35
-
***Dual-Model AI Engine**: Employs two distinct **Random Forest Classifiers** for robust detection. One model analyzes a comprehensive set of URL-based features, while the second incorporates the raw URL string itself for enhanced accuracy.
36
-
***Automated Classification**: URLs are classified as "phishing" if found in a verified database of **58 lakh URLs** from PhishTank and IEEE Dataport, or as "suspected" if newly identified and flagged by the AI.
37
-
***Scalable Architecture**: Built on a modular, pipeline-based system using **Django**, **PostgreSQL**, and **Celery** to handle asynchronous tasks and large data volumes.
38
-
***Automated Reporting**: Processes user requests submitted via a single input or Excel file list and emails a detailed, zipped report with all relevant attributes.
34
+
-**Proactive Domain Identification**: Utilizes fuzzy permutation with the `dnstwist` library to actively find potential phishing domains before they are used in attacks.
35
+
-**Dual-Model AI Engine**: Employs two distinct **Random Forest Classifiers** for robust detection. One model analyzes a comprehensive set of URL-based features, while the second incorporates the raw URL string itself for enhanced accuracy.
36
+
-**Automated Classification**: URLs are classified as "phishing" if found in a verified database of **58 lakh URLs** from PhishTank and IEEE Dataport, or as "suspected" if newly identified and flagged by the AI.
37
+
-**Scalable Architecture**: Built on a modular, pipeline-based system using **Django**, **PostgreSQL**, and **Celery** to handle asynchronous tasks and large data volumes.
38
+
-**Automated Reporting**: Processes user requests submitted via a single input or Excel file list and emails a detailed, zipped report with all relevant attributes.
@@ -56,12 +56,15 @@ Given this urgent and persistent threat, there is a critical need for a scalable
56
56
Our solution employs a multi-stage strategy that combines proactive domain identification with a sophisticated, dual-model AI engine.
57
57
58
58
#### Domain Identification
59
+
59
60
Our primary method for identifying new threats is **Fuzzy Permutation**, which uses the `dnstwist` library to generate and perform active DNS resolution on intentional misspellings and other permutations of legitimate CSE domains.
60
61
61
62
#### AI Classification
63
+
62
64
We use a pipeline of two distinct **Random Forest Classifiers** for robust detection.
63
-
***Model 1** analyzes a comprehensive set of **structural URL features**.
64
-
***Model 2** incorporates the **raw URL string** itself, processed via a TF-IDF Vectorizer, in addition to the structural features, which we found significantly improved accuracy.
65
+
66
+
-**Model 1** analyzes a comprehensive set of **structural URL features**.
67
+
-**Model 2** incorporates the **raw URL string** itself, processed via a TF-IDF Vectorizer, in addition to the structural features, which we found significantly improved accuracy.
65
68
66
69
The final classification is determined by averaging the confidence scores of both models. If one model's output is false while the other's is true, the output from Model 2 takes precedence due to its higher accuracy.
67
70
@@ -72,9 +75,10 @@ The final classification is determined by averaging the confidence scores of bot
72
75
Our system is built as a modular, full-stack application. The backend is a **Django** framework that manages all application logic and database operations. It handles asynchronous tasks using a **Celery** task queue with a **Redis** backend. User requests are stored in a **PostgreSQL** database.
73
76
74
77
The system pipeline is designed for efficiency and scalability:
75
-
***Request Ingestion**: User requests (single URL or Excel file list) are submitted via the frontend and pushed to a Celery task queue.
76
-
***Task Processing**: Celery workers pick up tasks from the queue to perform domain identification and AI classification.
77
-
***Reporting**: A separate Celery task generates zipped reports with domain details and screenshots, which are then emailed to the user.
78
+
79
+
-**Request Ingestion**: User requests (single URL or Excel file list) are submitted via the frontend and pushed to a Celery task queue.
80
+
-**Task Processing**: Celery workers pick up tasks from the queue to perform domain identification and AI classification.
81
+
-**Reporting**: A separate Celery task generates zipped reports with domain details and screenshots, which are then emailed to the user.
78
82
79
83
<center>
80
84
{% include figure.liquid loading="eager" path="/assets/img/phishnet/request_process_thread.png" class="img-fluid rounded z-depth-1" %}
@@ -89,26 +93,28 @@ The system pipeline is designed for efficiency and scalability:
89
93
Our AI models were trained on a substantial, custom-built dataset of **58 lakh URLs**. The project includes performance data for both models.
90
94
91
95
#### Model 1 (Structural Features Only)
92
-
***Accuracy**: 87%
93
-
***Confusion Matrix**:
94
-
||**Predicted Not Phishing**|**Predicted Phishing**|
95
-
| :--- | :--- | :--- |
96
-
|**Actual Not Phishing**| 4,787 | 533 |
97
-
|**Actual Phishing**| 677 | 3,489 |
98
-
***Key Feature Importances**: `entropyURL` (0.184), `averageSubdomainLength` (0.137), and `entropyDomain` (0.122).
96
+
97
+
-**Accuracy**: 87%
98
+
-**Confusion Matrix**:
99
+
||**Predicted Not Phishing**|**Predicted Phishing**|
100
+
| :--- | :--- | :--- |
101
+
|**Actual Not Phishing**| 4,787 | 533 |
102
+
|**Actual Phishing**| 677 | 3,489 |
103
+
-**Key Feature Importances**: `entropyURL` (0.184), `averageSubdomainLength` (0.137), and `entropyDomain` (0.122).
99
104
100
105
<center>
101
106
{% include figure.liquid loading="eager" path="/assets/img/phishnet/without_url_repeatedDigitsInURL_repeatedDigitsInSubdomain_cse_confusion_matrix.png" class="img-fluid rounded z-depth-1" %}
102
107
</center>
103
108
104
109
#### Model 2 (Raw URL + Structural Features)
105
-
***Accuracy**: 95%
106
-
***Confusion Matrix**:
107
-
||**Predicted Not Phishing**|**Predicted Phishing**|
108
-
| :--- | :--- | :--- |
109
-
|**Actual Not Phishing**| 6,363 | 286 |
110
-
|**Actual Phishing**| 313 | 5,098 |
111
-
***Key Feature Importances**: Character n-grams from the URL string, such as `s:/` (0.0196) and `tp:/` (0.0172).
110
+
111
+
-**Accuracy**: 95%
112
+
-**Confusion Matrix**:
113
+
||**Predicted Not Phishing**|**Predicted Phishing**|
114
+
| :--- | :--- | :--- |
115
+
|**Actual Not Phishing**| 6,363 | 286 |
116
+
|**Actual Phishing**| 313 | 5,098 |
117
+
-**Key Feature Importances**: Character n-grams from the URL string, such as `s:/` (0.0196) and `tp:/` (0.0172).
112
118
113
119
<center>
114
120
{% include figure.liquid loading="eager" path="/assets/img/phishnet/without_repeatedDigitsInURL_repeatedDigitsInSubdomain_cse_confusion_matrix.png" class="img-fluid rounded z-depth-1" %}
@@ -119,8 +125,9 @@ Our AI models were trained on a substantial, custom-built dataset of **58 lakh U
119
125
### Findings, Limitations, and Future Improvements
120
126
121
127
Our findings show that the dual-model approach significantly enhances detection accuracy. However, the current solution has a few limitations:
122
-
***Subdomain Detection**: The solution does not actively find phishing subdomains.
123
-
***Historical Data Analysis**: We have not yet integrated crawl indices for full historical analysis.
124
-
***Model Complexity**: Due to initial resource constraints, we used a Random Forest model, and we plan to explore more advanced models like **LSTMs**.
128
+
129
+
-**Subdomain Detection**: The solution does not actively find phishing subdomains.
130
+
-**Historical Data Analysis**: We have not yet integrated crawl indices for full historical analysis.
131
+
-**Model Complexity**: Due to initial resource constraints, we used a Random Forest model, and we plan to explore more advanced models like **LSTMs**.
125
132
126
133
Our roadmap includes plans to address these limitations by integrating the `subfinder` module for subdomain discovery and `cdx_toolkit` for historical analysis.
0 commit comments